Nearly Automated Left Ventricular Long Axis Tracking on Real Time Three-Dimensional Echocardiographic Data F Veronesi 1,2 , C Corsi 1 , EG Caiani 2 , C Lamberti 1 1 Dipartimento di Elettronica, Informatica e Sistemistica, Università di Bologna, Italy 2 Dipartimento di Bioingegneria, Politecnico di Milano, Italy Abstract The measurement of left ventricular (LV) long axis length (LAL) is an integral part of echocardiographic evaluation of LV volume, the most important determinant of LV systolic function. LAL measurements from 2D echocardiography (2DE) are highly dependent on the ability to obtain non-foreshortened LV images. Real-time 3D echocardiography (RT3DE) could potentially overcome the effects of long-axis foreshortening, but LAL measurements on RT3DE data are currently based on manual analysis and are time-consuming. We developed and tested a nearly automated method based on optical flow techniques for the measurement of the LV LAL throughout the cardiac cycle from RT3DE data. Results of comparisons on 10 patients with manual tracing on RT3DE data showed good agreement and no significant bias (r=.99, bias=-1.8mm). The proposed method allowed fast and accurate quantification of the LAL throughout the cardiac cycle with minimal user interaction and short computational time. 1. Introduction Longitudinal LV shortening defined as the difference between the long-axis length (LAL) at end diastole (ED) and at any other instant in the cardiac cycle, is an important parameter in the evaluation of left ventricular (LV) systolic function [1-2]. Moreover, an incorrect estimate of LAL could affect the measurements of crucial clinical parameters, such as LV volume, which is obtained from these estimates using geometric modelling. However, conventional two-dimensional echocardio- graphic (2DE) measurements of the LAL are subjective and time-consuming and rely on the ability to obtain non- foreshortened apical long axis views [3]. Real-time 3D echocardiography (RT3DE) is an emerging imaging technique that allows fast acquisition of volumetric datasets from a single transthoracic acoustic window. This new technology provides more complete information on left ventricular (LV) anatomy and function, which has recently triggered research protocols aimed at its testing and validation. By using advanced image processing techniques, it is possible to obtain from RT3DE data not only qualitative 3D visual information on chamber shape, function and wall motion [4], but also quantitative clinically useful parameters of LV function [5]. A recent paper [6] has highlighted the importance of selecting anatomically correct, non-foreshortened apical cross-sections from the RT3DE data for accurate estimates of LV mass. The major cause of the known underestimation of LV mass by 2DE was shown to be apical foreshortening [6]. However, in this study, the selection of the non-foreshortened apical views was performed off-line by manual analysis of RT3DE data. Because of the time consuming nature of this procedure, it was limited to end-diastolic (ED) and end-systolic (ES) frames and thus did not provide information on LAL changes throughout the cardiac cycle. We hypothesized that LV long axis could be automatically identified throughout the cardiac cycle from the RT3DE, and thus could provide the basis for fast, automated quantification of LV longitudinal shortening. Accordingly, the aim of this study was to develop a nearly automated method, based on optical flow techniques [7], to detect frame-by-frame the LV long-axis and measure its length. The automated LAL measurements were validated by comparisons with manual frame-by-frame measurements performed on the same RT3DE datasets. 2. Methods 2.1. Data acquisition Ten patients (age 52±12 years) with transthoracic acoustic windows that allowed adequate endocardial visualization without contrast enhancement were studied. RT3DE imaging was performed using a commercial ultrasound scanner (SONOS 7500, Philips) equipped with a fully sampled matrix array transducer (X4, 2-4 MHz) operating in the harmonic mode. RT3DE datasets were acquired using the wide-angled mode, wherein four wedge-shaped sub-volumes (93°x21°) were acquired over 0276-6547/05 $20.00 © 2005 IEEE 5 Computers in Cardiology 2005;32:5-8.